Sample Size Determination for Multivariate Performance Analysis with Complex Designs
The literature of multiple comparison methods addresses the problem of ranking treatment groups from best to worst. However, there is no clear indication of how to deal with the information from pairwise multiple comparisons, particularly in the case of blocking (or stratification) or in the case of multivariate response variables. In the present paper we take three methods into consideration to produce a performance ranking of C treatments under study. By means of a simulation study, it is possible to calculate the percentages of correct classifications of the compared methods and study their performances. The proposed simulation study also allows us to determine the minimum sample size useful for detecting performance differences among treatments.
KeywordsGlobal Score Complex Design Score Vector Multivariate Statistical Model Sample Size Determination
Unable to display preview. Download preview PDF.
The authors wish to thank the University of Padova (CPDA088513/08 and CPDA092350/09) and the Italian Ministry for University and Research (2008WKHJPK 002) for providing the financial support for this research.
- Bonnini, S., L. Corain, A. Cordellina, A. Crestana, R. Musci, and L. Salmaso (2009). A novel global performance score with application to the evaluation of new detergents. In Statistical Methods for the Evaluation of Educational Services and Quality of Products (M. Bini, P. Montanari, D. Piccolo and L. Salmaso eds). Heidelberg: Physica-Verlag, (to appear).Google Scholar